Sensorimotor cortex reinforcement learning of 2-joint virtual arm reaching (Neymotin et al. 2013)

 Download zip file   Auto-launch 
Help downloading and running models
Accession:150245
"... We developed a model of sensory and motor neocortex consisting of 704 spiking model-neurons. Sensory and motor populations included excitatory cells and two types of interneurons. Neurons were interconnected with AMPA/NMDA, and GABAA synapses. We trained our model using spike-timing-dependent reinforcement learning to control a 2-joint virtual arm to reach to a fixed target. ... "
Reference:
1 . Neymotin SA, Chadderdon GL, Kerr CC, Francis JT, Lytton WW (2013) Reinforcement learning of 2-joint virtual arm reaching in a computer model of sensorimotor cortex Neural Computation 25(12):3263-93 [PubMed]
Model Information (Click on a link to find other models with that property)
Model Type: Realistic Network;
Brain Region(s)/Organism:
Cell Type(s): Neocortex V1 pyramidal corticothalamic L6 cell; Neocortex U1 pyramidal intratelencephalic L2-5 cell; Neocortex V1 interneuron basket PV cell; Neocortex fast spiking (FS) interneuron; Neocortex spiking regular (RS) neuron; Neocortex spiking low threshold (LTS) neuron;
Channel(s):
Gap Junctions:
Receptor(s): GabaA; AMPA; NMDA;
Gene(s):
Transmitter(s): Gaba; Glutamate;
Simulation Environment: NEURON;
Model Concept(s): Synaptic Plasticity; Learning; Reinforcement Learning; STDP; Reward-modulated STDP; Sensory processing;
Implementer(s): Neymotin, Sam [samn at neurosim.downstate.edu]; Chadderdon, George [gchadder3 at gmail.com];
Search NeuronDB for information about:  Neocortex V1 pyramidal corticothalamic L6 cell; Neocortex V1 interneuron basket PV cell; Neocortex U1 pyramidal intratelencephalic L2-5 cell; GabaA; AMPA; NMDA; Gaba; Glutamate;
/
a2dmodeldb
readme.html
drspk.mod *
infot.mod *
intf6_.mod *
misc.mod *
nstim.mod *
stats.mod *
vecst.mod *
arm.hoc
basestdp.hoc
col.hoc
colors.hoc *
declist.hoc *
decmat.hoc *
decnqs.hoc *
decvec.hoc *
default.hoc *
drline.hoc *
filtutils.hoc *
geom.hoc
grvec.hoc *
hinton.hoc *
infot.hoc *
init.hoc
labels.hoc *
misc.h *
mosinit.hoc
network.hoc
nload.hoc
nqs.hoc *
nqsnet.hoc *
nrnoc.hoc *
params.hoc
python.hoc
pywrap.hoc *
run.hoc
samutils.hoc *
screenshot.png
sense.hoc *
setup.hoc *
simctrl.hoc *
stats.hoc *
stim.hoc
syncode.hoc *
trainedplast.nqs
units.hoc *
xgetargs.hoc *
                            
// $Id: filtutils.hoc,v 1.11 2011/11/07 21:37:48 samn Exp $ 

print "Loading filtutils.hoc..."

//* mkgauss(vector,average,standard-dev)
proc mkgauss () { local i,x,sz,mu,sd localobj vin
  vin=$o1 sz=vin.size mu=$2 sd=$3
  for vtr(&x,vin,&i) vin.x(i) = exp( -(x-mu)^2 / (2*sd*sd) )
  vin.mul( 1 / (sd*sqrt(2*PI)) )
}

//* mktriangwin(vec,size - should be odd,[skip the wraparound])
proc mktriangwin () { local i,j,sz localobj vin
  vin=$o1 vin.resize($2)
  vin.x(int($2/2))=1
  j=1 sz=1/(vin.size/2)
  for (i=int($2/2)-1;i>=0;i-=1) {
    vin.x(i)=j
    j-=sz
  }
  j=1
  for i=int($2/2)+1,vin.size-1 {
    vin.x(i)=j
    j-=sz
  }
  vin.div(vin.sum)
  if(numarg()>2) return
  vin.wraparound(vin.size)
}

//* mkgaussfilt(vec,stdev[,vx])
proc mkgaussfilt () { local sd,minx,maxx,dx,a localobj vin,vx
  vin=$o1 sd=$2
  if(numarg()>2) {vx=$o3 vin.resize(0) vin.copy(vx)}
  mkgauss(vin,0,sd)
  vin.wraparound(vin.size)
  vin.div(vin.sum)
  dealloc(a)
}

//* dofilt(vsignal,vwindow) - filters with convlv
proc dofilt () { local a,i localobj vsig,vwin,vtmp
  a=allocvecs(vtmp) vsig=$o1 vwin=$o2 sz=vsig.size
  vtmp.convlv(vsig,vwin)
  vsig.copy(vtmp)
  vsig.resize(sz) // make sure size doesn't change
  dealloc(a)  
}

//* triangfilt(vin,filtsize) - run a triangle filter
proc triangfilt () { local a localobj vin,vwin
  vin=$o1
  a=allocvecs(vwin)
  mktriangwin(vwin,$2) // make the window
  dofilt(vin,vwin) // do the filtering
  dealloc(a)
}

//* boxfilt(vin,filtsize) - run a box(moving average) filter
proc boxfilt () { local a localobj vin,vwin
  vin=$o1
  a=allocvecs(vwin)
  {vwin.resize($2) vwin.fill(1) vwin.div(vwin.size)} // make the window
  dofilt(vin,vwin) // do the filtering
  dealloc(a)
}

//* gaussfilt(vin,stdev,vx) - run a gaussian filter - vx is x-values used to make gaussian
proc gaussfilt () {  local a,sd localobj vin,vwin,vx
  vin=$o1 sd=$2 vx=$o3
  a=allocvecs(vwin)
  mkgaussfilt(vwin,sd,vx) // make the window
  dofilt(vin,vwin) // do the filtering
  dealloc(a)
}

//* myfilt(code,vec) - code:0=gauss,1=triangle,2=box
proc myfilt () { local a localobj vx
  if($1==0) {
    a=allocvecs(vx)
    vx.indgen(-3,3,.03)
    gaussfilt($o2,stdg,vx)
  } else if($1==1) {
    triangfilt($o2,winsz)
  } else if($1==2) {
    boxfilt($o2,winsz)
  }
}

//* resample(vec,new size) - resample a vec to new size using linear interpolation
proc resample(){ local newsz,idxdest,idxsrc,val,fctr,frac localobj vtmp
  {vtmp=new Vector($2) fctr=$o1.size/$2  vtmp.x(0)=$o1.x(0) idxsrc=fctr}
  for(idxdest=1;idxdest<$2-1;idxdest+=1){
    idxsrc = idxdest * fctr
    frac = idxsrc - int(idxsrc)
    idxsrc = int(idxsrc)
    if(idxsrc+1>=$o1.size){
      vtmp.x(idxdest) = $o1.x(idxsrc)
      continue
    }
    val = (1-frac) * $o1.x(idxsrc) + frac * $o1.x(idxsrc+1)
    vtmp.x(idxdest) = val
  }
  {vtmp.x($2-1)=$o1.x($o1.size-1) $o1.resize($2) $o1.copy(vtmp)}
}

//* bandfilt(datavector,low frequency,high frequency[,window size for bandpass filter])
// bandpass filter using via FFT convolution, $o1 will be modified
func bandfilt () { local a,idx,hz,lohz,hihz,wsz localobj v1,vwin,v2
  if(!name_declared("INSTALLED_myfft")){printf("bandfilt ERRA: myfft.mod not installed!\n") return 0}
  {v1=$o1  lohz=$2  hihz=$3}
  if(numarg()>3)wsz=$4 else wsz=1025
  a=allocvecs(vwin,v2)
  vwin.resize(wsz)
  vwin.bpwin(lohz/sampr,hihz/sampr) //make bandpass windowed sinc
  vwin.wraparound(wsz) //wrap around for FFT convolution
  v2.convlv(v1,vwin)
  v1.copy(v2)
  dealloc(a)
  return 1
}

//* multfilt(inputvector,outputvector,vector of lowfrequencies,vector of highfrequencies)
//get superposition of multiple frequency bands, each pair in $o3,$o4 should be considered a band
func multfilt () { local a,i localobj vtmp,vin,vout,vlo,vhi
  if(!name_declared("INSTALLED_myfft")) {printf("multfilt ERRA: myfft.mod not installed!\n") return 0}
  a=allocvecs(vtmp)
  vin=$o1 vout=$o2 vlo=$o3 vhi=$o4
  for i=0,vlo.size-1 {
    vtmp.copy(vin)
    bandfilt(vtmp,vlo.x(i),vhi.x(i))
    if(i==0)vout.copy(vtmp) else vout.add(vtmp)
  }
  vout.resize(vin.size)
  dealloc(a)  
  return 1
}

//* nrnpsd(vector,samplingrate) - calculates PSD and returns as an NQS object
// uses NEURON spctrm function
obfunc nrnpsd () { local sampr localobj vec,nqp
  vec=$o1 sampr=$2
  nqp=new NQS("f","pow")
  nqp.v[1].spctrm(vec)
  nqp.v.indgen(0,sampr/2,(sampr/2)/nqp.v[1].size)
  nqp.v.resize(nqp.v[1].size)
  return nqp
}

//* mkprefftwin(outputvec,windowtype[,windowsize]) - make window for multiplying before running fft
proc mkprefftwin () { local pref,sz localobj vpre
  vpre=$o1 pref=$2
  if(numarg()>2) sz=$3 else sz=vpre.size
  vpre.resize(sz)
  if(pref==1) {
    vpre.blackmanwin()
  } else if(pref==2) {
    vpre.hanningwin()
  } else if(pref==3) {
    vpre.hammingwin()
  } else {printf("mkprefftwin ERRA: unknown window type!\n") return}
}

//* getspecnq(vec,sampr[,specty,prefilt])
obfunc getspecnq () { local sampr,specty,pref,a localobj vec,vpre,v2,nq
  vec=$o1 sampr=$2 nq=nil
  if(numarg()>2) specty=$3 else specty=0
  if(numarg()>3) pref=$4 else pref=0
  if(pref) { // pre-fft filtering
    a=allocvecs(vpre,v2)
    vpre.resize(vec.size)
    mkprefftwin(vpre,pref)
    v2.copy(vec)
    v2.mul(vpre)
    vec=v2
  }
  if(specty==0) nq=pypmtm(vec,sampr)
  if(specty==1) nq=pypsd(vec,sampr)
  if(specty==2) nq=matpmtm(vec,sampr)
  if(specty==3) nq=nrnpsd(vec,sampr)
  if(specty==4) nq=pybspow(vec,sampr)
  if(pref) dealloc(a)
  return nq
}


Loading data, please wait...